beyond network selection: exploiting access network hetero- - - PowerPoint PPT Presentation

▶
beyond network selection exploiting access network hetero
SMART_READER_LITE
LIVE PREVIEW

beyond network selection: exploiting access network hetero- - - PowerPoint PPT Presentation

Nov 22, 2023 469 likes •727 views

2nd ACM Conference on Information-Centric Networking . Klaus M. Schneider, Udo R. Krieger October 2, 2015 University of Bamberg, Germany 0 beyond network selection: exploiting access network hetero- geneity with named data networking 1

slide-1
SLIDE 1

beyond network selection: exploiting access network hetero- geneity with named data networking

2nd ACM Conference on Information-Centric Networking .

Klaus M. Schneider, Udo R. Krieger October 2, 2015

University of Bamberg, Germany

slide-2
SLIDE 2

two trends: content & mobility

Source: Cisco VNI Mobile, 2015 1

slide-3
SLIDE 3

multihomed terminal scenario

Content

Mobile Client

Ad-hoc WiFi L TE WiMAX 2

slide-4
SLIDE 4

characteristics of the scenario

Common Observations

  • 1. Different Application Requirements
  • 2. Different Access Network Characteristics
  • 3. Different Cost Factors

Goal: Doing better than IP Network Selection

3

slide-5
SLIDE 5

solution: stateful & adaptive forwarding!

Source: Van Jacobson et al. - Networking Named Content (2009) 4

slide-6
SLIDE 6

system architecture & design

5

slide-7
SLIDE 7

system architecture – interface estimation

  • 1. Passive Monitoring
  • 2. Active Probing

→ Moving Average

6

slide-8
SLIDE 8

system architecture – forwarding strategies

7

slide-9
SLIDE 9

strategies - best interface first

8

slide-10
SLIDE 10

strategies - packet striping

9

slide-11
SLIDE 11

strategies - parallel transmission

10

slide-12
SLIDE 12

Implementation & Evaluation

11

slide-13
SLIDE 13

delay estimator

= TCP RTT Estimator → Exponential Moving Average

12

slide-14
SLIDE 14

loss estimator

No discrete loss values! Own Loss Estimator Design! Based on Simple Moving Average

13

slide-15
SLIDE 15

loss estimator

No discrete loss values! Own Loss Estimator Design! Based on Simple Moving Average

13

slide-16
SLIDE 16

loss estimator – measurement results

Client R2 R1 R3 Backbone Server Face 257 F a c e 2 5 6 F a c e 2 5 8

10 20 30 40 5 10 15 20

Time [s] Measured Packet loss [%]

Face 256 (10%) 257 (20%) 258 (30%)

Uniform Packet Loss

14

slide-17
SLIDE 17

bandwidth estimator

Naive bandwidth probing of unused paths → Too much overhead! ฀ 1. Passive Bandwidth Estimation

25 50 75 100 125 5 10 15

Time [s] Bandwidth [KByte/s]

Face 256 (1000 KB/s) 257 (500 KB/s) 258 (100 KB/s) 15

slide-18
SLIDE 18

bandwidth estimator – 2. burst estimation

200 400 600 800 1000 5 10 15

Time [s] Bandwidth [KByte/s]

Face 256 (1000 KB/s) 257 (500 KB/s) 258 (100 KB/s) 16

slide-19
SLIDE 19

lowest cost strategy

Chooses lowest cost face that satisfies all requirements! ฀ Requirements: maxloss, maxdelay, minbw ฀ Cost: Ordinal attribute Example: maxdelay=500ms, maxloss=10%, prio=delay

Client R2 R1 R3 Backbone Server Face 257 F a c e 2 5 6 F a c e 2 5 8

17

slide-20
SLIDE 20

lowest cost strategy – measurement results

250 500 750 1000 10 20 30 40

Time [s] Data Rate [Pkts/s]

Face 256 257 258

10 sec: Loss deterioration of path 257 20 sec: Delay deterioration of path 256 30 sec: Delay recovery 40 sec: Loss recovery

18

slide-21
SLIDE 21

madm strategy

Multiple Attributes + Chooses Highest Value Interface More flexible requirements:

QoS Metric A QoE QoS Metric B QoE QoS Metric C QoE

Two thresholds: Min and Max Example: delay=20ms-50ms, cost1=10-20, cost2=2-5

19

slide-22
SLIDE 22

madm strategy: mobile traffic limit scenario

App1: loss=0%-40% App2: loss=0%-40%, cost=50%-70% cost = tconsumed tlimit (1)

100 200 300 10 20 30 40 50

Time [s] Packet Rate [Pkts/s]

Face 256 257 Type inData

  • utInt

20

slide-23
SLIDE 23

selective parallel strategy

Scenario: Real-time communications! SP = LCS + Selective Flooding

  • 1. Req. not met: Flood
  • 2. Req. not met: Use best 2 paths
  • 0.25

0.50 0.75 1.00 BestRoute LCS SP−Best2 SP−Flood Broadcast

Forwarding Strategy

  • Perc. inside Req.
  • 200

400 600 800 BestRoute LCS SP−Best2 SP−Flood Broadcast

Forwarding Strategy Sent Interests [Pkts/s]

21

slide-24
SLIDE 24

summary

Adapting NDN to Wireless Multihomed Devices ฀ Interface Estimation (QoS Awareness) ฀ New Adaptive Forwarding Strategies Future Work

  • 1. Details of Parallel Forwarding Strategies
  • 2. Realistic Evaluation Scenarios
  • 3. Congestion-Control

22

slide-25
SLIDE 25

the end

Thank you for your attention!

Klaus Schneider

klaus.schneider@uni-bamberg.de klaus@cs.arizona.edu

23